17 strategies run side by side on live market data. Every rule they check, every token they refuse and every trade they close is written down and published here — the losses with the rest.
The strategies on the board are test configurations and the numbers they produce are provisional — calibration is ongoing.
One point per closed trade. These trades never shared a wallet, so the line is what a single 12 ETH book would have done taking every one of them — 12 ETH because the most this desk ever had deployed at once was 10.74. Every exit is marked at the quoted price, never at a fill.
1,538 of 2,755 trades closed down, most of them badly. The book is carried by the right-hand tail — the few that were left alone to run.
| Token | Strategy | Held | ETH | Result |
|---|---|---|---|---|
| XCOINS 10 Sept, 11:15 UTC | Launch L2 TP5x | 11m | +0.1115 | +557.4% |
| ARHD 09 Sept, 14:41 UTC | Launch L2 TP5x | 20m | +0.1073 | +536.3% |
| FAT 08 Sept, 08:32 UTC | Launch L2 TP5x | 7m | +0.0962 | +481.2% |
| MARIO 10 Sept, 14:37 UTC | Launch L3 Hold60 | 22m | +0.0859 | +429.4% |
| Hoper 09 Sept, 08:23 UTC | Launch L2 TP5x | 11m | +0.0845 | +422.5% |
| haMSTR 08 Sept, 10:04 UTC | Launch L2 TP5x | 13m | +0.0822 | +411.2% |
| ELIZABAO 08 Sept, 16:13 UTC | Launch L2 TP5x | 18m | +0.0809 | +404.6% |
| p(doom) 09 Sept, 16:43 UTC | Launch L2 TP5x | 13m | +0.0806 | +403.1% |
The live decision log, straight from the engine. Identical consecutive refusals are collapsed — a standing no is one line, not two hundred.
| Token | Strategy | Held | ETH | Result |
|---|---|---|---|---|
| MARIO 10 Sept, 16:01 UTC | Launch L0 | 25m | +0.0479 | +239.4% |
| MARIOBROS 10 Sept, 16:01 UTC | Launch L0 | 25m | −0.0036 | −17.9% |
| SUSHISTOCK 10 Sept, 16:01 UTC | Kayle | 3.1h | −0.0082 | −81.9% |
| MARIO 10 Sept, 16:01 UTC | Mcap M0 | 25m | −0.0059 | −29.7% |
| FLY 10 Sept, 16:01 UTC | Mcap M0 | 25m | +0.0328 | +164.1% |
| XPONS 10 Sept, 16:01 UTC | Launch L5 Cap30 | 21m | −0.0170 | −84.8% |
| ZFORGE 10 Sept, 16:01 UTC | Mcap M3 TP1.5x | 25m | +0.0137 | +68.6% |
| MARIO 10 Sept, 16:01 UTC | Launch L3 Hold60 | 25m | +0.0465 | +232.6% |
| ZFORGE 10 Sept, 16:01 UTC | Mcap M0 | 25m | +0.0137 | +68.6% |
| TENEV 10 Sept, 16:01 UTC | Mcap M3 TP1.5x | 25m | +0.0033 | +16.3% |
Every token the desk actually buys gets a full write-up — supply, holders, the deployer's history, how the launch went. The ones already published are here.
An agent earns its keep by what it declines. Every test below was run against a real token in the last 96 hours, and almost all of them ended the same way — a rule was not met, so nothing happened.
every test the engine ran against a real token
a rule was not met — the token's, the chart's, or the desk's own spending cap
every rule met at once
entered and exited inside the same window
The same market, the same window, different rules. The ones in the red stay on the board — that is the point of running them all at once.
| # | Strategy | Trades | Win rate | Avg | Best | Median hold | Result ETH | |
|---|---|---|---|---|---|---|---|---|
| 01 | sim | 32 | 47% | +20% | +269% | 1.0h | +0.092 | |
| 02 | sim | 35 | 37% | +18% | +176% | 12.0h | +0.626 | |
| 03 | sim | 35 | 49% | +13% | +247% | 20m | +0.088 | |
| 04 | sim | 2 | 50% | +10% | +20% | 17m | +0.002 | |
| 05 | sim | 506 | 47% | +7% | +353% | 20m | +0.295 | |
| 06 | sim | 496 | 40% | +6% | +429% | 1.0h | +0.320 | |
| 07 | sim | 501 | 43% | +6% | +338% | 20m | +0.209 | |
| 08 | sim | 511 | 45% | +5% | +557% | 21m | +0.060 | |
| 09 | sim | 23 | 43% | +1% | +63% | 2.5h | +0.033 | |
| 10 | sim | 36 | 39% | +1% | +81% | 2.5h | +0.038 | |
| 11 | sim | 28 | 50% | −0% | +69% | 1.0d | −0.012 | |
| 12 | sim | 163 | 44% | −2% | +194% | 1.0h | −0.108 | |
| 13 | sim | 75 | 36% | −3% | +333% | 3.0h | −0.020 | |
| 14 | sim | 160 | 38% | −6% | +301% | 1.0h | −0.302 | |
| 15 | sim | 151 | 29% | −10% | +388% | 4.1h | −0.293 | |
| 16 | sim | 1 | 0% | −59% | −59% | 23m | −0.012 | |
| 17 | sim | 0 | — | — | — | — | 0.000 |
Everything above came out of one loop, running every minute: read the chain, take the token apart, decide, size it, and write down why. The strategies are the dial settings — the machine is what turns them.
Every new pool on Robinhood Chain, screened on volume, liquidity and market cap before anything else runs. Hundreds of tokens a day narrowed to the few worth a second look.
Supply concentration, bundler and sniper wallets, the deployer's other launches, transfer taxes, honeypot checks — the things that decide whether a chart is even real.
Tokens that clear the rules get a full report: the dip, the holders, the launch, the liquidity, and a verdict in plain English. Published here as they are finished.
Entry, take profit, stop loss and holding time are fixed before the trade opens, and the exit is marked to what actually came back — never to the quote that opened it.
17 strategies compete on the same market at the same time. One market, one window, different rules — which is the only honest way to find out which rules are worth having.
294 refusals for every signal. The agent does nothing unless every condition is met at once, and the reason it declined is written down either way.
Every decision on this page came from the engine's own log, with the measurement that produced it. Nothing is summarised after the fact, and nothing that lost money has been left out.
Define the conditions, deploy an agent, and watch it work in the same detail you just read.